A Recovery Approach for Real-Time Database Based on Transaction Fusion

In a real-time database, correctness of transaction depends not only on the correct implementation, but also depends on the effectiveness of time. Traditional database recovery strategies are adopted to undo all malicious transactions and affected transactions, and then redo the affected transactions. This time-consuming recovery process is not acceptable for real-time databases. Sometimes a transaction may have already exceeded its implementation period, and thus lose the meaning of recovery. By assessing the repair necessity for the affected transaction, a recovery approach is proposed to decide whether to recovery, and then fuse malicious transactions and valuable affected transactions. By this way it can significantly reduce the required number of the transactions needed to undo and redo, avoid worthless recovery operations, and reduce the system I/O and log scanning. So the recovery can be finished in an efficient time.